| The development of urban road tunnel rapidly,but many issues related to the research is still in its infancy stages,the intelligent road tunnel monitoring system is a necessary condition to ensure normal and efficient operation of the road tunnel,and this paper studies the control of the tunnel drainage system which is an important part of road tunnel monitoring system.The formation of objects of drainage is strong non-linear and spatial variability, It's difficult to understand clearly that its complex changing mechanism,the system model and nonlinear system identification method is used,but from theory to methods there are still many places need to improve.Currently the BP network is commonly used in identification of nonlinear systems,but BP neural network is a kind of static neural network,it is only the implementation of static non-linear one-to-one corresponding mapping,not suitable for real-time identification of dynamic system.On the basis of analyzing dynamic traffic flow in urban road tunnel, this paper puts forward the dynamic neural network model of the traffic flow forecasting in urban road tunnel,the mode based on the Elman network,with functions of the state memory,time series model established by Elman network is a autoregressive moving average model.Output of the model depends not only on its past and present input,but also on the output of the past.It can highly improve the forecast accuracy.Atthe same time,the formation of drainage object is an uncertain phenomenon, the initial state can not determine the unknown state,and these uncertainties include random,fuzzy and incomplete information(gray) and the imprecise nature of information processing.Objective uncertainties mainly include rainfall,topography, surface conditions,such as the randomness and fuzziness.These objective uncertainty led to an uncertainty flow analysis and simulation of subjective uncertainty,such as the calculation model,the selected parameters,calculation assumptions,calculation simplify,the calculation icon,information description,and measurement accuracy. And compared with the deterministic method,fuzzy logic showed greater scientific superiority in describing the characteristics of things and our understanding of the relationship which is not clear.Because the majority samples in flow forecasting Neural network models small samples,it is not enough to learn the rules of large flow samples,it often resulting in low flow forecasting.To address this problem,consider the classification of pre-flow to increase the representation of large volume samples.We need to establish a neural network model for each category,on the basis of these model,to establish the choice of control model based on Fuzzy inferences,the new samples through the fuzzy rule base to determine discriminates classification into different networks forecasting model.Compare with traditional drainage control system,the system's pump is running up to about 14%energy saving,energy-saving effect is very obvious.Urban highway tunnel drainage neuron-fuzzy control system has a strong feasibility and practical value of application;it has an important practical significance for the design of current road tunnel drainage control system. |
PDF Full Text Request